Scholarly article on topic 'Mapping properties to monitor forests: Landholder response to a large environmental registration program in the Brazilian Amazon'

Mapping properties to monitor forests: Landholder response to a large environmental registration program in the Brazilian Amazon Academic research paper on "Law"

CC BY
0
0
Share paper
Academic journal
Land Use Policy
OECD Field of science
Keywords
{"Environmental monitoring" / "Land tenure" / "Property size" / Deforestation / Amazon / Policy}

Abstract of research paper on Law, author of scientific article — Jessica L’Roe, Lisa Rausch, Jacob Munger, Holly K. Gibbs

Abstract Across the tropics, development banks and conservation donors are investing millions in property mapping and registration projects to improve accountability for deforestation. An evaluation of the effectiveness and accuracy of existing environmental registries is crucial to assure the success of future efforts. This study presents an evaluation of deforestation and registration behavior in response to one of the largest of these property registration programs to date — the Rural Environmental Registry (CAR) in the Amazonian state of Pará. From late 2007 to 2013, approximately 100,000 properties covering 30 million hectares of self-declared claims were entered in this digital registry. We used fixed effects regression models and property level data to assess how registration influenced deforestation on different sizes of properties. Registration had little impact on deforestation behavior, with the exception of a significant reduction on “smallholder” properties in the size range of 100–300ha. We link this reduction to interacting incentives from forest protection and land regularization policies and suggest that desire to strengthen land claims motivates these landholders’ response to the environmental registry. We also present evidence that some landholders may be registering incomplete or inaccurate parcels into the self-declared system to strategically benefit from policy incentives. Our results for smallholder properties indicate that environmental registries may have potential to facilitate reductions in deforestation if combined with a favorable combination of incentives. However, in places where land tenure is still being negotiated, the utility of environmental registries for forest policy enforcement and research may be limited without ongoing investment to resolve uncertainty around land claims.

Academic research paper on topic "Mapping properties to monitor forests: Landholder response to a large environmental registration program in the Brazilian Amazon"

Contents lists available at ScienceDirect

Land Use Policy

journal homepage www.elsevier.com/locate/landusepol

Land Use Policy

Mapping properties to monitor forests: Landholder response to a large environmental registration program in the Brazilian Amazon

Jessica L'Roe3'*, Lisa Rauschb, Jacob Mungerb, Holly K. Gibbsa b

a Department of Geography, University of Wisconsin-Madison, 550 North Park Street, Madison, WI53706, USA

b Center for Sustainability and the Global Environment (SAGE), University of Wisconsin-Madison, 1710 University Avenue, Madison, WI 53726, USA

CrossMark

ARTICLE INFO

Article history: Received 18 August 2015 Received in revised form 19 May 2016 Accepted 25 May 2016

Keywords:

Environmental monitoring

Land tenure

Property size

Deforestation

Amazon

Policy

ABSTRACT

Across the tropics, development banks and conservation donors are investing millions in property mapping and registration projects to improve accountability for deforestation. An evaluation of the effectiveness and accuracy of existing environmental registries is crucial to assure the success of future efforts. This study presents an evaluation of deforestation and registration behavior in response to one of the largest of these property registration programs to date — the Rural Environmental Registry (CAR) in the Amazonian state of Para. From late 2007 to 2013, approximately 100,000 properties covering 30 million hectares of self-declared claims were entered in this digital registry. We used fixed effects regression models and property level data to assess how registration influenced deforestation on different sizes of properties. Registration had little impact on deforestation behavior, with the exception of a significant reduction on "smallholder" properties in the size range of 100-300 ha. We link this reduction to interacting incentives from forest protection and land regularization policies and suggest that desire to strengthen land claims motivates these landholders' response to the environmental registry. We also present evidence that some landholders may be registering incomplete or inaccurate parcels into the self-declared system to strategically benefit from policy incentives. Our results for smallholder properties indicate that environmental registries may have potential to facilitate reductions in deforestation if combined with a favorable combination of incentives. However, in places where land tenure is still being negotiated, the utility of environmental registries for forest policy enforcement and research may be limited without ongoing investment to resolve uncertainty around land claims.

© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license

(http://creativecommons.org/licenses/by/4.0/).

1. Introduction

1.1. The importance and challenge of clarifying land ownership

Many of the world's most biodiverse and carbon-rich forests are located in places where land tenure is uncertain. It is difficult to design programs that reduce deforestation without knowing who is making land use decisions, and landscape level conservation initiatives require information about private lands beyond protected areas. Reliable property maps are now a requirement for many REDD projects because property boundary data help signal whom to reward for ecosystem services (Larson et al., 2013; Naughton-Treves and Wendland, 2014; Duchelle et al., 2014). Property maps also aid environmental monitoring by indicating whom to penal-

* Corresponding author. E-mail addresses: Jessica.LRoe@wisc.edu (J. L'Roe), llrausch@wisc.edu (L. Rausch), munger@wisc.edu (J. Munger), hkgibbs@wisc.edu (H.K. Gibbs).

ize for land use violations. Beyond facilitating the allocation of carrots and sticks, property maps can improve understanding of the drivers of land use change by illuminating different patterns among different types of actors (Geoghegan et al., 1998; Liverman and Cuesta, 2008). For these reasons and others, conservation and development organizations, from The Nature Conservancy to the Inter-American Development Bank, are investing millions in large scale land registration programs in Brazil, Colombia, Haiti, Peru, Guatemala, Rwanda, Ecuador and many other tropical countries.

Despite the potential benefits, clarifying land ownership is not without risks. For forests, these include hastened deforestation to establish claims (e.g. De Oliveira, 2008), conflict induced deforestation accompanying land reform (e.g. Alston et al., 2000), and heavier investments in forest-displacing agriculture under greater tenure security (e.g. Liscow, 2013). Risks for people include loss of informal collective access, exacerbated conflict, and capture of land by elites (Araujo et al., 2009; Rajao, 2013; Paulino, 2014). Many of these risks are associated with the act of establishing or changing rights to land (e.g. land reform and land titling programs). Because programs

http://dx.doi.org/10.1016/j.landusepol.2016.05.029

0264-8377/© 2016 The Authors. Published by Elsevier Ltd. This is an open access article underthe CC BY license (http://creativecommons.org/licenses/by/4.0/).

that allocate land rights can be expensive and politically fraught (Deininger and Feder, 2009), and because outcomes for forests can be mixed (Angelsen and Kaimowitz, 1999; Robinson et al., 2014), groups concerned with forest management are experimenting with land registration programs that simply map existing claims. Rather than alter the security of land rights, the goal of this type of environmental land registry is to reveal and systematize information about land users for the purposes of monitoring and planning (Bennett et al., 2013).

Programs to reform land rights have received much scholarly attention, but few studies have evaluated the outcomes of programs to map land claims in environmental registries, partly because these systems are relatively new and evolving (Gignoux et al., 2013; but see Rajao et al., 2012; Azevedo et al., 2014). Environmental registries are intended to reduce deforestation by facilitating monitoring and enforcement of environmental policies. However, if most land is not registered, if there are conflicting incentives, or when property data are not accessible or reliable, then registries will not necessarily reduce deforestation. There is also a risk that land users may attempt to subvert the system by undertaking clearing prior to registering, by registering only parts of their land, or by interpreting their registration as a permit to clear (Chomitz, 2007; Rajao et al., 2012; Azevedo and Saito, 2013). This paper provides empirical evidence to aid the design of ongoing land registration initiatives and to elucidate discussions about the role of different actors and incentives that affect Amazon deforestation. We address the following questions about environmental registries: first, under what circumstances does registration cause people to reduce deforestation, or perversely, to increase it? next, what are the tradeoffs between information quantity and information quality when land registrations are declaratory?, and finally, what are the implications of mapping properties without resolving land rights if landowner-ship is still being negotiated, as is often the case at development frontiers? To engage these questions, this study evaluates one of the largest environmental registration programs in the tropics.

1.2. Background: land registration in Para's Rural Environmental Registry

Environmental registration initiatives in Brazil are globally important because they cover millions of hectares of forested land and are serving as examples for programs in other countries. Brazil's Forest Code requires land users to maintain forest on a certain proportion of their property—up to 80% for most of the Amazon biome (Soares-Filho et al., 2014). To facilitate monitoring and enforcement of this policy, the Brazilian government is developing a System of Rural Environmental Registries (Portuguese acronym: SICAR) with the aim of mapping and digitizing all rural properties in the country. The new national system draws from several antecedent state systems, each with somewhat different procedures and objectives, beginning with the System for Environmental Licensing (Portuguese acronym: SLAPR) in the State of Mato Grosso in the early 1990s (Azevedo and Saito, 2013). Expectations for these programs have been high among the international environmental community and within Brazil (Fearnside, 2003; Chomitz, 2007). For example, a recent study quoted a senior Brazilian government official: "[with GIS], deforestation in these areas is going to have a name and a surname. This fact certainly leads to a sensible increase in the governance capability of the environmental agencies in the Amazon states" (Rajao et al., 2012). We evaluate deforestation and registration behavior in response to the largest of the state programs, Para's Rural Environmental Registry (Portuguese acronym: CAR), which began in 2008 and by 2013 comprised over 100,000 property boundary registrations covering over 30 million hectares of Amazon Forest (Fig. 1).

There are several characteristics of the environmental registry in Para that affect its impacts and how it might apply to other regions. Land distribution in Para is highly skewed; small properties (here a "small" property can exceed 100 ha) abut ranches of several thousand hectares or more due to waves of agrarian reform settlements and colonization projects targeting different actors (Pacheco, 2009). Para is famous for land conflicts and properties continue to be claimed, expropriated, and fought over—sometimes violently (Schmink, 1982; Fearnside, 2001; Wright and Wolford, 2003; Simmons, 2005). Forged titles are common, official titles are rare, and the ability to make and defend claims to land is at the forefront of people's minds, especially in frontier regions (Oliveira, 2013; Campbell, 2014; Reydon et al., 2015). Faced with the challenge of informal and contested land rights in much of the state (one study estimated as much as 53% as of 2008) (Barreto et al., 2008), the designers of Para's CAR program adopted a two-phased system. In the first phase, the goal was coverage: as many claims as possible would be mapped by land users, who essentially drew the boundaries of properties on high resolution imagery, aided by GIS technicians deployed across the state (SIMLAM, 2008; Benatti and Fischer, 2011). This provisional stage epitomizes the goal of trying to simply map claims without altering or adjudicating land rights. A "definitive" CAR license could later be obtained after a second phase in which property boundaries and ownership were externally verified and owners submitted a plan for complying with environmental laws (SIMLAM, 2008). Five years after the start of the program, as of November 2013, only 2% of the CAR registrations in Para had advanced to "definitive" licenses. The other 98% of registrations were "provisional" CAR licenses, automatically generated in response to self-declared boundaries entered in an online system.

1.3. Theory: incentives, perceptions, and property size

A policy's impacts come partly from its design and objectives, but also from the way people respond according to their own aims and agendas. We examine property size patterns in Para's environmental registry to provide clues about landholder motivations when they registered their holdings.

Why would landholders choose to make their properties more visible to environmental monitoring? Both the government and private sector have offered an array of incentives to encourage registration, including access to certain kinds of credit and markets1 (Azevedo et al., 2014; Gibbs et al., 2015). Another strong incentive for mapping a property into a government system in this region is to bolster the strength and legitimacy of land claims, even though a provisional CAR license expressly did not signify any change in the legal status of land claims in Para (Para State Decree N° 1,148/2008). Nonetheless, field-based studies have described a common sentiment that land policy in this region is in an uncertain stage of rapid formation and as such, claimants have been hustling to gain a favorable position under policies deemed likely in the future; mapped boundaries in a government registry may gain legal status later (Campbell, 2014). Moreover, for properties below a certain size, enrolling in the environmental registry was a required step in a concurrent but less widespread program for obtaining formal land titles through Brazil's land regularization program, Terra Legal (Duchelle et al., 2014). Terra Legal aims to do what CAR does not - it is program designed to help smallholders acquire land titles. Despite policy-makers' intentions to keep the processes of mapping and claiming separate, landholders may reasonably perceive the processes to be overlapping.

1 Registration in the CAR was technically mandatory in Para beginning in 2008, but

sanctions for non-compliance were seldom, if ever, applied (Azevedo et al., 2014).

Fig. 1. Distribution of land in Para's Rural Environmental Registry (CAR). Dark gray shows areas registered in the CAR as of November 2013. White areas are not yet registered but technically eligible. Registrations overlap designated protected areas, both in places where this is legal (some sustainable use areas) and where it is not (strict protection, military, and indigenous areas).

Why would registration induce changes in deforestation behavior? Landholders may reduce deforestation on registered properties because of concern about increased visibility to monitoring and enforcement, or in response to incentives associated with being visibly compliant. Concerns about enforcement should not apply to only a specific property size class, but policy incentives can be size specific. For example, in the revised Forest Code, clearing that happened prior to July 2008 on small properties is forgiven and smallholders can use all of their remaining forest area for compensation programs, while any deficiencies in the legal reserve on larger properties must be corrected through reforestation or compensation (Soares-Filho et al., 2014). Overlaying the Forest Code provisions, regularizing claims on government land2 is contingent on environmental compliance only for properties above a certain size (Benatti and Fischer, 2011). The suites of regulations

and incentives applying to different property size classes allow us to explore whether some sets of incentives are associated with stronger responses to the environmental registry.

The self-declared boundaries in the environmental registry may in many instances be the most official record of a property that is publicly available. As such, we also investigate whether incentives lead registrants to declare their property boundaries strategically, in order to fall in a size class with more advantageous policies. For example, in federal and state land regularization laws, the process for obtaining title is easier for smallholders3 — on site inspection may not be required for small properties while approval from congress is required for large properties (Pacheco and Benatti, 2015). Apart from its intended role in facilitating deforestation reduction by enhancing monitoring and enforcement, the registry

2 Much ofthe land in Para is still officially held by the state or federal government; this is the default until landholders make it through the process of regularizing their land claims and obtaining formal title.

3 For small properties (<~100ha), land is donated by the government to the claimant, medium-sized properties can be purchased directly from the government at a discounted rate, and large properties (>~1500ha) must be purchased at public auction (Pacheco and Benatti, 2015).

is also meant to aid policy design by clarifying debates about the extent to which smallholders vs. largeholders bear responsibility for current deforestation (Godar et al., 2012, 2014; Michalski et al., 2010; Richards and Vanwey, 2015). Property boundaries that reveal the locations of large and small actors over a wide region could also support efforts to understand how interactions between smallholders and largeholders drive deforestation, including displacement of smallholder settlers by incoming ranchers (Fearnside, 2001; Pacheco, 2012) and conflict between smallholders and largeholders during struggles to expropriate and redistribute land (Alston et al., 2000; De Oliveira, 2008; Aldrich et al., 2012). The usefulness of the registry in this respect is contingent on the assumption that registrations reflect the actual sizes of landholdings.

1.4. Objectives

In the remainder of this paper, we evaluate the initial outcomes associated with Para's Rural Environmental Registry. We focus first on understanding the degree to which land registration has led to near-term changes in deforestation behavior. We test for effects in property size groups subject to different policy incentives, isolate the effects of registration from background trends within these groups, and check for perverse increases in deforestation prior to registration. Next, we assess whether the ease and accessibility of a self-declared registration system comes with tradeoffs in information quality. We inspect histograms of registration size to investigate whether registrations are declared in strategic and potentially misleading sizes, and we measure the extent of overlapping registrations where claims are not resolved. We link behavior to policy and suggest that key aspects of landholder response to the CAR thus far may be motivated by desire to stake and strengthen land claims, even though CAR was specifically designed to be an environmental registry and not a land titling program. We conclude by discussing broader implications of these findings for land registration initiatives.

2. Methods

2.1. Defining the dataset

Our analysis is based on registrations in the CAR database as of November 2013 (n=101,654)4 (SIMLAM-PA, 2013). In its raw form, the registry contains many instances of identical and overlapping properties. We removed all exact duplicates5 from the data set so that properties would not be double-counted. Summary statistics were calculated over the remaining set of 98,058 properties. We used PRODES data for land cover and deforestation (INPE, 2014). For the models assessing impacts of registration on deforestation behavior, we excluded properties with no remaining primary forest at the beginning of the study period where continued deforestation was no longer possible (n = 20,373). We also excluded properties smaller than five ha to mitigate resolution issues when assigning deforestation to individual properties (n = 2421). We excluded properties in locations that were not fully covered in the preliminary release of the 2014 PRODES forest cover

4 This was the most recent version available at the time of analysis. Additional properties continued to be registered after this date, and in 2014 the new Federal system began to replace the state system.

5 Duplicates were defined as properties whose spatial footprint was >95% identical. Incases where multiple registrations were issued to the same landowner for the same area, we kept the observation with the earliest registration date. In cases where registrations were issued at multiple dates to multiple entities for the same land area, we assumed these may have been cases of sales and kept the earliest registration date. We removed registrations of outer boundaries of land reform settlement areas but kept any individual properties registered within settlements.

map, the most recent available at time of analysis (n = 3261). Finally, to avoid double-counting deforestation and to circumvent model identification problems from assigning multiple registration dates to one place, we excluded properties when more than 5% of their area overlapped with another registered property (n = 34,596).6 In combination, these restrictions reduced the number in the model sample to 45,942 properties. We present model results from this entire sample as well as from two samples with additional restrictions. Additional restrictions hone in on areas with greater potential for influence from land titling policies: properties whose majorities were outside of land reform settlements (INCRA, 2014) and properties whose spatial footprints were >95% identical to properties that were also enrolled in the Terra Legal land titling program (Terra Legal, 2015).

We defined size classes based on a standard definition in Brazilian land policy (Federal Law 8,629/1993) used in other studies from this region (Azevedo et al., 2014): properties less than four fiscal modules7 are "small," properties between four and 15 fiscal modules are "medium," and properties greater than 15 fiscal modules are "large." Module sizes vary with municipality (between 50 and 75 ha in rural Para) and we used the appropriate municipal definition when classifying registrations. However, as an approximation for the reader: small properties are less than roughly 300 ha, medium properties range up to about 1000 ha, and large properties are generally greater than 1000 ha. There are many more small properties than medium and large properties, and we further divided this group into "small" properties of less than 100 ha and "semi-small" properties between 100 ha and 4 modules. One hundred hectares is often the maximum parcel size allocated in land reform settlement projects, and four modules is the legal cut-off when defining a small family farm for economic purposes (Federal Law 11,326/2006). Thus our small category captures most land reform settlers and what would traditionally be thought of as "smallholders," and our semi-small category could be settlers who have acquired additional land, wealthier "smallholder" colonists, or partially declared properties of larger landholders.

2.2. Modeling the impact of registration on deforestation behavior

To assess deforestation outcomes related to property mapping, we modeled the effects of registration timing on the amount of deforestation occurring on a property in a given year.8 Essentially, we assess how registration changes the behavior of those who enroll in the program by comparing the rates of deforestation on registered properties to the rates of deforestation on properties that were not yet registered but that eventually registered by November 2013. This approach ensures that we measure effects associated with the act of registration and not effects associated with differences in the kinds of landholders that choose to participate in the system or the kind of land that gets enrolled (i.e. selection bias). With the panel data, we used fixed effects linear regression to control for the average levels of deforestation on each property due to time constant factors like distance to market or

6 Because this group was so large, we performed robustness checks without this restriction and with higher overlap thresholds. The model results were not sensitive to variations in this restriction.

7 Land and property laws in Brazil often refer to the concept of a 'fiscal module'. This is a size unit that is defined for each municipality according to the amount of land required for the primary economic activities of that region (Federal Decree 84,685/1980).

8 We used the listed date of registration in the system — provisional CAR licenses were granted almost instantaneously in Para. Registration dates were aggregated

into years beginning in August and running through July in order to synch more closely with the time window considered in the maps of annual deforestation released by the PRODES satellite monitoring program (e.g. 2010=August 2009-July 2010).

geographic characteristics, and we included additional year level fixed effects to control for background temporal trends occurring across the sample due to factors like fluctuations in commodity prices. The combination of property level information with fixed effects regression helps to isolate changes in deforestation behavior due to registration from the suite of other dynamics affecting forest clearing. Models followed the two structures shown below, for property i in time t.

Deforit = yi +yt x (ifregisteredbeforeorduringyeart)it + eit (1)

Deforit = yi + yt + x (ifregisteredinyeart + 1)it +

x (ifregisteredinyeart)it +

x (ifalreadyregisteredbeforeyeart)it + eit (2)

We examined annual deforestation on a property (Deforit) using both a binary indicator of whether or not any deforestation occurred in a given year and a continuous indicator for the amount of deforestation occurring in a given year.9 The variables yi and yt are the property and year fixed effects, respectively. Standard errors are clustered at the municipality level to account for spatial autocorrelation of properties. For Eq. (1), the 'treatment' variable is a binary indicator for whether a property was registered in a given year, including properties that were already registered in prior years. We ran this model over the time period from August 2009 to July 2013, when the bulk of registrations occurred.For Eq. (2), the 'treatment' variable is broken into three categories, where we separated the effects of the year immediately prior to registration, the year of registration, and the years after registration, relative to the reference level of over a year prior to registration. We ran the Eq. (2) model for August 2008 to July 2012, the years with sufficient properties in each category of registration timing (Table 1). This model allowed us to check for increases in deforestation immediately prior to registering that could potentially have been induced by knowledge that properties would be monitored after registration. To allow for heterogeneous registration effects and background trends according to property size, we ran separate models for our four size-classes of properties.

2.3. The search for evidence of strategic registration

We examined the distribution of registrations using a histogram plotting the frequency of properties of different sizes. Peaks indicate relatively common sizes. We would expect a high frequency of properties allocated at standard sizes in colonization and resettlement programs (e.g. 100 ha), with clustering around these standard sizes due to imprecise surveying and other technical factors. In contrast, much denser distributions on only one side of a peak suggest that land is acquired, or is simply being registered, with an eye toward property size thresholds related to policy incentives. For example, if up to 100 ha of land can be claimed for no cost, one might expect more 98 ha parcels than 102 ha parcels to be registered. Landholders might declare only 99 ha of a larger piece they use, or might register a 190 ha parcel as two 95 ha units. We test

9 We used the inverse hyperbolic sine transformation for our continuous measure ofthe amount of deforestation. This behaves almost identically to the more common log transformation, except that it is also defined for values of zero (Burbidge et al., 1988). This allows for handling the frequent cases of no deforestation in a given year similar to a binary measure, but also varies with extent of deforestation when it occurs. We also modeled the proportion of remaining forest cleared in a given year; patterns were basically the same as the binary and continuous measures. For all deforestation measures, we used linear models for ease of interpretation with panel data.

Small Semi-small

Fig. 2. Falling background trends for deforestation by size class. Averages are shown forthe 45,942 properties in the deforestation models, which exclude properties with no forest remaining in 2007.

for strategic registration by investigating instances where properties were registered disproportionately frequently at sizes just below policy thresholds. We calculated the ratio between properties that registered just below and those just above threshold values along a scale based on increments of one hundred hectares, and along a scale based on increments of fiscal modules.10 Based on our review of pertinent land and forest laws, Table 2 summarizes which of these incremental thresholds corresponds with actual policy thresholds and incentives. If people register to strategically qualify for the benefits of one of these policies, we would expect disproportionate registration below the corresponding size thresholds. Though the environmental registry is not directly connected to several of the policies in Table 2, we include them because landholders may think of the registry as an opportunity to officially delineate boundaries in a way that may help them in other arenas (e.g., future land regularization efforts).

3. Results

3.1. Summary statistics for deforestation trends and landholdings

As of late 2013, roughly half of the eligible land in Pará was registered in the CAR system. Small properties accounted for the majority of registrations, yet most of the registered land area and even more of the registered forest area was held in larger properties (Table 3). Across all size classes, the amount of deforestation declined from 2007 to 2011 and then stabilized (Fig. 2). Properties in the largest size class still clear more on average than properties in the smallest class, but absolute deforestation rates declined most steeply on large properties-thus, the mean annual amount of deforestation on small and large properties has become more comparable over the past decade. In fact, while deforestation at the beginning of the study period in 2007 was much more likely to occur on larger properties, by 2014, deforestation was just as likely to be found on small properties (Table 3). These general trends are related to processes other than CAR registration. The panel models seek to identify additional reductions in deforestation associated with enrolling in the environmental registry, beyond these significant background trends.

10 We tested within a range of ±5 ha. In the absence of threshold effects, the ratio of properties just under vs. just over a certain size should be close to 1:1 within a narrow window of 10 ha.

Table 1

Timing of registration for each size class in the model sample of properties.

Year3 Small Semi-small Medium Large Total

2008 31 14 10 16 71

2009 95 40 67 82 284

2010 1716 1275 1334 834 5159

2011 3684 1975 1018 595 7272

2012 5895 2646 1293 666 10,500

2013 8420 2988 1393 726 13,527

2014 6638 1519 650 322 9129

Total 26,479 10,457 5765 3241 45,942

a Year follows the PRODES calendar and includes the previous August through July of the indicated year (e.g. August 2007-July 2008 = "2008"). Italicized years (2008 and 2009) were excluded from impact models because there were very few registrations. Our dataset only contains registrations from the first 4 months of 2014 (August-November 2013).

Table 2

Summary of legal thresholds.

Threshold

Summary of policies relevant to property size thresholds

Relevant law

1 fiscal module (50-75 ha) 100 ha

4 fiscal modules (200-300 ha)

300 ha 500 ha

15 fiscal modules(750-1125 ha)

1500ha 2500 ha

Below this amount, federal lands are donated to claimants. Below this amount, state lands are donated to claimants. Below this amount, forest cleared prior to 2008 is forgiven and claims for federal lands are not subject to onsite inspection if the property remains free of additional violations.

In 2010 the state made it easier for properties below this

size to obtain a CAR license by lowering the amount of

required data and increasing the number of institutions

authorized to register boundaries.

Above this amount, property claims require an economic

use plan and are subject to verification by an outside

agency.

Below this amount, land claims can be regularized through the Terra Legal program and purchased at discounted rates; above this amount federal land must be purchased at public auction.

Above this amount, approval from the state assembly is required for purchase of state lands. Above this amount, congressional approval is required for purchase of both federal and state lands.

Federal Land Regularization Law (11,952/2009) Para Land Regularization Law (7,289/2009) Federal Forest Code (12,651/2012); Federal Decree on Land Regularization (6,992/2009)

Normative Instruction Amending Para CAR Law (37/2010)

Para Decree on Land Regularization (2,135/2010)

Federal Land Regularization Law (11,952/2009)

Para Decree on Land Regularization (2,135/2010)

Federal Land Regularization Law (11,952/2009); Para Decree on Land Regularization (2,135/2010)

Table 3

Relative importance3 of different size classes in the environmental registry.

Portion of Portion of Portion of Portion of Portion of registered deforestation in 2014b (%)

registrations (%) registered land registered registered

(%) primary forest deforestation in

(%) 2007b (%)

Small 63.5 10.6 6.1 13.2 29.1

Semi-sm 21.1 11.0 7.5 15.4 25.1

Medium 10.2 20.0 18.2 24.9 20.6

Large 5.2 58.4 68.1 46.4 25.2

a Percentages are based on the larger set of 98,058 properties, after duplicates were removed but before restricting sample to properties that still contain forest. However, trends are similar when calculated for the restricted model sample and all values differ by less than five percentage points.

b The portion of deforestation that falls within each size category, out of the total detected by PRODES for that year within the area registered in the CAR by November 2013 (INPE, 2014).

3.2. Effects of registration on deforestation among different property size classes

Registration had little effect on deforestation behavior overall. This result is sensitive to the time window used in the model. Including deforestation from years prior to 2010 leads to a larger overall effect. Our focus on the years in which the bulk of properties registered is conservative. However, responses to registration varied by property size-class. Among semi-small properties, reductions were modest but statistically significant. On these properties, registration reduced the probability of any deforestation by nine percent and reduced the extent of deforestation by 5.3%, relative to unregistered properties (Table 4). This pattern persisted when we

excluded properties that overlapped land reform settlements. No significant effects were evident for any other size classes.11

The modest but statistically significant effect of registration for semi-smallholders was greater among the subset of registrants known to be seeking land titles. We found that the deforestation-reducing effect was about three times larger on CAR properties that

11 We also tested for the effect in properties at the small end of the medium size range (those between fourand seven modules) as a robustness checkthat the difference in registration effects represents a discontinuity at the legal threshold defining smallholders and not a gradual difference corresponding to property size. Even among the smaller properties in the medium category, registration did not cause a decrease in deforestation behavior in the way that it did for the properties under the legal smallholder cut-off.

Table 4

Effect3 of registration in CAR system on annual deforestation by size class.

Overall Small Semi-small Medium Large

Full Sample of CAR Effect on probability of defor(% change) -2.6 -2.3 -9.2* 4.3 3.2

registrations Effect on extent of clearing (% change) -1.9 -1.4 -5.3*** 1.2 -1.3

n 45,942 26,479 10,457 5765 3241

Set of CAR registrations Effect on probability of defor(% change) -4.4 -7.4 -8.7t 2.8 3.2

excluding land reform Effect on extent of clearing (% change) -1.9t -2.1 -5.2** 0.6 -1.3

settlement areasb n 29,571 13,680 7745 4905 3241

Set of CAR registrations Effect on probability of defor(% change) -9.4 4.4 -28.6* 3.2 N/A

in Terra Legal titling Effect on extent of clearing (% change) -4.7 2 -13.8* -2.3 N/A

processc n 2328 1156 822 330 N/A

a Significance levels indicated by f: p<0.1; *: p<0.05; **: p<0.01; ***: p<0.001. Results are from models using a binary outcome for deforestation (probability) and a continuous transformed measure of deforestation (extent). Percent change was calculated relative to the modeled deforestation outcome for unregistered properties. Results are based on analysis time window from August 2009 - July 2013.

b This is the subset of properties that fall outside areas identified as INCRA land reform settlements. Land reform settlements are subject to different and additional land governance policies and procedures.

c This is the subset of the properties registered in the CAR whose spatial locations align with properties in the Terra Legal database; i.e. those known to be actively seeking title. Large properties are not eligible for titling through this program.

Table 5

Effecta of registration timing on deforestation.'

Overall Small Semi-Small Medium Large

Year before registration 0% 1% 0% -4% 0%

Year of registration -3% -1% -7%* -1% -5%

After registration year -2% 3% -11%* -2% 1%

a *: p < 0.05.

b Results from the model using the continuous (transformed) measure of deforestation. Results are expressed as the percent change from the reference level of deforestation over a year prior to registration.

were also part of the Terra Legal land titling program, but again, only in the semi-smallholder class. In this program, during the mortgage period when landholders are purchasing the property from the state, environmental infractions can trigger cancellation of the process. For the subset of semi-small smallholders seeking land titles in Terra Legal, the chance of deforestation drops by almost 30% after property registration.

3.3. Strategic registration and ambiguous ownership

To test for strategic clearing before entering the system, we explored how deforestation changed for properties about to register, those in the midst of registering, and those already registered in a previous year, rather than making a simple before-and-after comparison (Eq. (2)). We find no significant increase in deforestation in the year immediately preceding registration for properties of any size class (Table 5). In concordance with earlier results, only semismall properties show a significant decrease in clearing behavior during the year of registration and in subsequent years.

Though we do not find evidence for strategic clearing behavior, we do find evidence for strategic registration. As expected, we see high frequencies of properties allocated at standard sizes in colonization and resettlement programs (e.g. 100 ha), with clustering around these standard sizes due to imprecise surveying and other technical factors. We also see places where there are denser distributions on only one side of a peak, suggesting that land is registered with an eye toward property size thresholds related to policy incentives. A histogram of registered property sizes (Fig. 3) reveals several locations where property size frequency stacks disproportionately below particular peaks. Thresholds are evident at 300 ha, 1125 ha,12 1500 ha, and 2500 ha. Visually, it is difficult to tell whether other peaks in the histogram exhibit disproportion-

12 75 ha is one of the most common fiscal module sizes, which means that for 45% of the properties in our sample, 1125 ha and 15 fiscal modules represent the same threshold, as do 300 ha and 4 fiscal modules. There are no municipalities in

ate stacking, but ratio tests revealed additional threshold effects (Table 6). Significantly more properties register just under13 1,2,4, and 15 fiscal modules, as well as 100, 200, 300, 500, and 1500 ha. Almost all of these sizes correspond to specific legal thresholds,14 which are largely about tenure regularization and not directly connected to the environmental registry (see Table 2).

Finally, we found a substantial amount of overlap among land registrations in the declaratory system. When intersecting the spatial dataset of registered property boundaries with itself, 27,991 (27.5%) of property registrations overlapped by more than 5% and less than 95% with another registration.15 These overlapping registrations may signal inaccurately drawn boundaries, contentious land claims, or accumulated versions of properties that were reregistered with different shapes for a variety of potential reasons.

4. Discussion

4.1. Has the environmental registry caused landholders to reduce deforestation?

Though deforestation fell in the region during our study period, most of this decline was not a direct result of the environmental registry. The majority of registrants did not substantially alter their clearing behavior immediately following (or preceding) registration, in our time window of analysis. Note that our study was designed to measure how registration affected individual landholder behavior, which is different than the effect of the program on deforestation overall.16 We can speak to the impact of registration on enrolled properties, but we cannot speak to deforestation and land claiming dynamics on land that was not enrolled. For

Para where a fiscal module is equal to 100 ha, so, for example, 1500 ha and 15 fiscal modules are never equivalent.

13 The only instance where significantly more properties registerjust over a threshold occurs in the small group around 900 ha. This is likely due to reasons apart from regularization and forest policies; we found no references to this size cut-off in our research.

14 There are no corresponding legal thresholds that we are aware of to explain the moderately higher proportions registering areas just under 2 fiscal modules and 200 ha unless there is intent to later split the properties.

15 Overlapping areas greater than 5% are not easily explainable by small mapping errors, and overlap less than 95% excludes cases of multiple registrations ofthe same property due to system errors and sales, or small individual properties registered within larger settlement boundaries.

16 Because registrations are the unit of analysis, changes in deforestation on small properties count as much as changes on large properties. This tells more about how the CAR is working as a suite of incentives (i.e. how many people have changed their behavior) and less about how much reduced deforestation the program has stimulated.

o o in

Area of CAR registration (axis is on log-scale)

Fig. 3. Histogram of registered property sizes. The distribution evinces distinct modes and propensity to stack below certain size thresholds.

Table 6

Proportion of properties just undera vs. just overa a range of size thresholds/

threshold (modules) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Ratioa 1.2 1.4 1.1 2.3 1.1 0.9 0.9 1.3 0.9 1.0 1.1 0.8 1.1 0.9 2.8 1.1

Countb 4,818 1,063 587 834 205 167 226 118 110 89 70 64 89 120 252 56

P-valc 0.00 0.00 0.22 0.00 0.49 0.16 0.57 0.07 0.28 0.88 0.74 0.16 0.65 0.44 0.00 0.70

threshold (hectares) 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600

Ratioa 1.2 0.9 1.9 1 1.2 0.9 0.9 0.8 0.6 0.9 0.8 1.4 1.0 0.9 1.7 1.3

Countb 9,919 1,044 697 215 330 101 78 69 67 161 89 49 24 30 83 23

P-valc 0.00 0.02 0.00 0.92 0.03 0.67 0.34 0.23 0.01 0.44 0.18 0.07 1.00 0.61 0.00 0.38

a The ratio of the properties in the five hectares just below a threshold compared to the five hectares just above a threshold. b The number of registrations within ±five hectares of a threshold.

c The probability that the proportion just under is equal to the proportion just over, given the number of properties within five ha of the threshold. Known legal thresholds are highlighted in boldface.

enrolled properties, a weak or neutral initial effect on deforestation is not surprising since the primary aim of the CAR program in Para thus far has been to gather boundary information to facilitate future monitoring and enforcement. Though in theory it made properties publicly visible, few organizations had full access to the CAR database, and it was rarely used for enforcement against illegal deforestation (possibly to avoid discouraging future registrations). If property holders did not feel that registration increased the likelihood of enforcement, they would have had little reason to change deforestation behavior. This may especially be the case for larger properties whose boundaries were more likely to be publicly documented by other means outside of the CAR (Rajao and Vurdubakis, 2013; Reydon et al., 2015) and thus were already vulnerable to enforcement associated with monitoring programs that preceded the CAR, such as the DETER satellite monitoring program (Assuncao et al., 2014; Arima et al., 2014). A report from the Amazon Environmental Research Institute (IPAM) found a similar result of little overall impact from Para's CAR, though they report larger reductions on small and medium properties and an increase in deforestation on large properties (Azevedo et al., 2014). This may be due in part to different time windows of analysis and in part to the use of year-by-year comparisons among registered and unreg-

istered properties rather than fixed effects models. Evaluations of the antecedent environmental licensing program (SLAPR) in the neighboring state of Mato Grosso also indicated that registration did not decrease deforstation and in some instances increased it (Rajao et al., 2012; Azevedo and Saito, 2013). We found no evidence for significant effects of increased deforestation before or after registration in Para.

Despite the lack of a strong overall effect on clearing behavior thus far, our results demonstrate that registration did lead to a decrease in deforestation for one particular size class: the semi-small properties (100 to ~300 ha). The discontinuity between behavior on semi-small properties and those in neighboring size classes suggests that the explanation is related to policies that treat these groups differently rather than inherent differences among large and small actors. The deforestation responses of different property size classes could be explained if some landholders enrolled in the environmental registry in hopes that it would facilitate land regularization. This is supported by the fact that the deforestation reduction effect is three times larger for the semismall properties known to be already involved in the Terra Legal

titling program.17 These properties may have reduced deforestation after registering because they fell in an unintended policy "sweet-spot" created by an interaction of land regularization and forest protection policies.

Semi-small properties are at the larger end of Brazil's legal definition for "small family farms.They are larger than typical land reform settlement allocations, but could be owned by wealthier land reform settlers who have acquired additional land. Semi-small properties could also be owned by settlers with historic claims on land that still officially belongs to the government, or semismall properties could belong to landholders with more extensive properties registered in smaller pieces for strategic reasons.18 Semi-small properties are large enough that landholders must pay to purchase the land from the government if they want to regularize their claims,19 though they pay a rate below market value and can pay it off over a 10-year period (Para State Decree 2,135/2010, Federal Decree 6,992/2009). The right to gain title might be cancelled if environmental infractions are detected on the property during this pay-down period (Para State Decree 2,135/2010). Environmental infractions can also trigger an external inspection of the property claim that is otherwise waived for properties in this size class (Federal Decree 6,992/2009). This means that, at least until a property is paid down, obtaining a title is facilitated by compliance with environmental policies. This does not apply to properties less than 100 ha; these are donated to claimants by the government and thus do not have a sensitive mortgage period, which may explain the lack of response in the smallest size class. Nor does it apply for large properties over ~1500ha that must go through other legal channels outside of the Terra Legal program to receive title. Medium-sized properties should still be subject to the government mortgage period, but we do not see evidence that registration affects deforestation behavior in that size class. Perhaps property owners in this class are wealthy enough to buy their land without a government mortgage, but we would not expect wealth to be discontinuous between semi-small properties and medium properties. An alternative explanation is that, for the medium-sized properties, the cost of coming into compliance may overwhelm any benefits that such compliance might confer on easing the titling process (Stickler et al., 2013). There is a discontinuous cost of compliance between semi-small and medium properties: in the 2012 Forest Code, areas cleared prior to mid-2008 do not need to be replanted or compensated on properties smaller than four fiscal modules (Law 12,651/2012). Environmental compliance is disproportionately more difficult for properties just over the size threshold for forgiveness. Medium properties must replant or pay to compensate any areas cleared beyond their designated Legal Reserve, which can be as much as 80% of the property. This means that the semi-small properties are in the unique position where environmental compliance both facilitates receiving a land title and does not incur heavy costs.

17 Far more properties are registered in the CAR than there are to date in the slower moving Terra Legal land titling program. However, the presence of the Terra Legal in the region may have significantly influenced landholder behavior, especially since a CAR was required to be eligible for regularization through Terra Legal. Properties not yet involved in the titling process might still have seen a CAR registration as a potential step toward strengthening property claims.

18 Although technically these properties would be ineligible for most smallholder benefits, which apply only to producers with small total landholdings, we are not aware of any measures that prevent families from registering areas up to particular limits in the names of various relatives and pooling their use.

19 In the absence of an official title, land in the Amazon falls in the category of"unas-signed public land" held by the government. Though it is technically unassigned, much has long been used and occupied. The 2009 Land Regularization Law establishes a procedure for occupants to purchase their land claims from the government (Oliveira 2013).

A reduction in deforestation on semi-small properties is encouraging, especially because smallholders can be hard to reach with market-based and command-and-control policies (Coudel et al., 2012), and smaller properties are becoming responsible for an increasing component of the deforestation in Para as deforestation rates decline more steeply on larger properties. However, while the semi-small properties represent 21% of registrations in Para, they contain only 7.5% of registered forested land. Many small properties contain no forest at all. At the end of 2013, half of the eligible land in the state was not yet registered to any size class. Thus, a substantial amount of unprotected primary forest is not yet captured within the registration system, and the vast majority of forest on registered properties does not fall in the size class that is currently incentivized to reduce deforestation following registration. Field-based studies in Para indicate that many landholders in Para are aware of legal thresholds, especially those related to titling and land regularization (Campbell, 2015).

4.2. Information quality under declaratory registration

Denser stacking of property sizes just below legal thresholds indicates that landholders may not be registering their entire or exact holdings. This could mean registering parts of their property and leaving the rest unregistered, or registering land in strategically-sized segments under different names, or claiming conveniently-sized pieces that may be unrelated to what they are actually using. Strategic registration could hamper the effectiveness of environmental monitoring. If people do not register their entire landholdings, then it is possible that future deforestation may be concentrated in unregistered places that avoid enforcement, resulting in spatial leakage of deforestation. If people split land into strategically-sized pieces under different names, this could either allow people who are not truly eligible to fall into the category where deforestation prior to 2008 does not need to be compensated, or allow people to keep a 'clean' property as a basis for sales while committing environmental infractions on other properties they are using (Gibbs et al., 2015). One of the objectives of the environmental cadastres in the Amazon is to reduce enforcement expenses by decreasing the need for costly site visits to remote areas (Rajao et al., 2012). However, if there are questions about the extent to which registered boundaries represent actual holdings, or if holdings are overlapping, site visits remain just as necessary as before. Evidence for strategic claim-sizes also raises a red flag concerning the use of CAR to understand the deforestation impacts of smallholders versus largeholders; it is not necessarily safe to assume that formally declared small properties actually belong to smallholders.

The ease of registering claims in the Para CAR has paid off in coverage; over 30,000,000 ha of land are now registered, at least provisionally. The CAR as it stands represents a dramatic increase in information and its creation has not led to perverse deforestation outcomes. Yet, jumping ahead to map claims without first resolving or legitimizing claims has risks. Use of the database for enforcement will be hampered by lack of certainty about the accuracy of the information it contains, and it is unlikely that the registry can have a large or continuing effect on deforestation without stronger connection to enforcement. It remains to be seen whether Para's initiative will prove a worthwhile investment for reducing deforestation. As the new federal CAR system (SICAR) comes on line, state and federal environmental agencies are beginning to move the 98% of registrations that made it only to the "provisional" phase in Para through the more stringent verification and environmental recovery planning (PRA) phases of the registration process. This will require all boundary overlaps to be resolved and calls for heavy investment in inspection procedures. In the meantime, care must be taken when using and interpreting data from what was meant

to be an initial phase.20 Field validations of the registry data should be undertaken to inform efforts to expand or replicate the program elsewhere.

4.3. Mapping for environmental monitoring or staking land claims

Strategic positioning with regard to land claims is predictable in a system where boundaries are essentially self-declared. Indeed, land holdings and land claims are notoriously chaotic in Brazil, and especially in the Amazon, where there have been competing and fraudulent claims for centuries (Brannstrom, 2001; Jepson, 2006; Bourguignon, 2013). However, it is somewhat surprising to see strategic positioning that is more about land claims than environmental policy in an environmental registry that was meant to be separate from questions of tenure security. Stacking beneath four fiscal modules could signal that owners were responding to the clause in the Forest Code that forgives historic clearing under that threshold. However, none of the other sizes where stacking occurred are mentioned in the Forest Code, while all the thresholds (including four fiscal modules) are meaningful in land regulariza-tion laws, specifically various criteria that grant an easier titling process. The decline in deforestation for semi-small properties was likely due to its perceived connection with titling. This bolsters the argument that landholders may have come to view the environmental registry as a hopeful step on a path toward a future land title, rather than as a tool that facilitates monitoring of their land use. The process of issuing land titles through Terra Legal and other channels has been criticized as being unacceptably slow and having negligible impact (Oliveira, 2013), but our findings suggest that landholders are paying attention to their government's efforts to resolve tenure issues in the Amazon and that the impacts of such programs may be magnified by other policies.

The link between the environmental registry and land regular-ization means that from the perspective of relevant landholders, the CAR might be perceived as part of a land regularization initiative with tenure incentives contingent on environmental performance. This is despite the fact that during the period under study, there was not yet a legal basis for CAR boundaries to be translated into legal land claims, and both the Para CAR program and the newer federal SICAR program explicitly stated that environmental registration was not intended to support land claims (Para State Decree N° 1,148/2008, Portal Brasil, 2015). Nevertheless, local perceptions and expectations may not be unfounded; in 2013, modifications to the state's Green Municipalities program opened the path to title for some properties registered with CAR in municipalities that had achieved certain environmental goals (Para State Decree 739/2013). Furthermore, there are currently discussions in Para about the potential to use the CAR database in future official land titling efforts (Monteiro, 2015). This turn of events would certainly legitimize registrants' hopes that the CAR would eventually become a step toward land titles. If this happens, our findings about strategic claiming around size thresholds and interactions between environmental and tenure policy incentives should become even more relevant.

5. Conclusion

Para's experience with its Rural Environmental Registry program offers several lessons for other environmental registration

20 For example, there are reports that some banks and vendors accept a provisionalCAR license as proofofenvironmental compliance in orderto meet industry policies to restrict loans to or sales from people who practice illegal deforestation (Azevedoet al., 2014).

initiatives. It is possible to map claims prior to resolving them, and the combination of few verification requirements, positive incentives to register, and an easy declaratory process facilitated a high volume of registrations, including among smallholders who can be under-represented in property mapping efforts. We show that some motivations for enrolling in the registry and reducing deforestation are tied to incentives from a parallel land policy as opposed to the incentives offered directly by the CAR. In particular, some are hoping that boundary maps from the CAR will facilitate the land titling processes. This may also explain registration's impact on deforestation for certain size classes, because the land titling process is also facilitated by avoiding environmental infractions. Our findings highlight the need to pay attention to the interaction of policies and incentives that can produce "sweet-spots,strategic leveraging, and other combinatory effects. Our results suggest that under strong incentives, environmental registries have the potential to reduce deforestation, but without such incentives, their effect on deforestation behavior is neutral. The registry's value for forest conservation depends primarily on the usefulness of the data for future enforcement of environmental regulations, not as a 'fix' in and of itself. This means the trade-off between volume of registrations and legitimacy of registrations must be taken seriously, along with the ways that parallel programs can affect registration incentives. Though self-declared registrations facilitate wider coverage, self-declaration of boundaries under differentiated policy incentives may encourage users to make claims that misrepresent their actual land use or landholdings. This may limit the degree to which the registry can facilitate environmental enforcement. In places where land tenure concerns rival environmental pressures, government and conservation sponsors will need to continue to invest in resolving uncertainty related to land ownership.

Acknowledgements

We gratefully acknowledge Lisa Naughton-Treves, Andrew L'Roe, the Gibbs Land Use Lab, the UW-Geography Biodiversity Interest Group, and three anonymous reviewers; their attentive input greatly improved this study. Jennifer Alix-Garcia and Fanny Moffette provided helpful discussion of analytical techniques. Funding provided by the Gordon and Betty Moore Foundation and Norwegian Agency for Development Cooperation's Department for Civil Society under the Norwegian Forest and Climate Initiative. All errors are our own.

References

Aldrich, S., Walker, R., Simmons, C., Caldas, M., Perz, S., 2012. Contentious land change in the Amazon's Arc of Deforestation. Ann. Assoc. Am. Geogr. 102, 103-128.

Alston, L.J., Libecap, G.D., Mueller, B., 2000. Land reform policies, the sources of violent conflict, and implications for deforestation in the Brazilian Amazon. J. Environ. Econ. Manage. 39,162-188.

Angelsen, A., Kaimowitz, D., 1999. Rethinking the causes of deforestation: lessons from economic models. World Bank Res. Obs. 14 (1), 73-98.

Araujo, C., Bonjean, C.A., Combes, J.L., Motel, P.C., Reis, E.J., 2009. Property rights and deforestation in the Brazilian Amazon. Ecol. Econ. 68 (8), 2461-2468.

Arima, E.Y., Barreto, P., Araujo, E., Soares-Filho, B., 2014. Public policies can reduce tropical deforestation: lessons and challenges from Brazil. Land Use Policy 41, 465-473.

Assuncäo, J., Gandour, C., Rocha, R., 2014. DETERring deforestation in the Brazilian Amazon: environmental monitoring and law enforcement. Working Paper, Available: http://hdl.handle.net/123456789/48637 (accessed 6.08.15).

Azevedo, A.A., Saito, C.H., 2013. O perfil dos desmatamentos em Mato Grosso, apos implementafäo do licenciamento ambiental em propriedades rurais. Cerne Lavras 19(1), 111-122.

Azevedo, A., Rajäo, R., Costa, M., Stabile, M.C., Alencar, A., Moutinho, P., 2014. Cadastro Ambiental Rural. In: Boletim Amazonia em Pauta. IPAM.

Barreto, P., Pinto, A., Brito, B., Hayashi, S., 2008. Quem e o dono da Amazonia? uma analise do recadastramento de imoveis rurais. Instituto do Homem e Meio Ambiente da Amazonia-IMAZON.

Benatti, J.H., Fischer, L.R.D.C., 2011. Land Use Regulations in the state of Pará, Brazil: an overview of its guidelines. Global Land Project: Workshop on Land Use Transitions in South America. Sao Paulo, Brazil.

Bennett, R., Van Gils, H., Zevenbergen, J., Lemmen, C., Wallace, J., 2013. Continuing to bridge the cadastral divide. Proceedings of the Annual World Bank Conference on Land and Poverty, 8-11.

Bourguignon, C., 2013. Leases, concessions, and authorizations. Searching for an alternative to the privatization of federal domain in Brazil. Land Use Policy 33, 100-110.

Brannstrom, C., 2001. Producing possession: labour, law and land on a Brazilian agricultural frontier, 1920-1945. Polit. Geogr. 20, 859-883.

Burbidge, J.B., Magee, L., Robb, A.L., 1988. Alternative transformations to handle extreme values of the dependent variable. J. Am. Stat. Assoc. 83,123-127.

Campbell, J.M., 2014. Speculative accumulation property-making in the Brazilian Amazon. J. Lat. Am. Caribb. Anthropol. 19, 237-259.

Campbell, J.M., 2015. The Land Question in Amazonia: Cadastral Knowledge and Ignorance in Brazil's Tenure Regularization Program. PoLAR: Political Legal Anth. Rev. 38(1), 147-167.

Chomitz, K.M., 2007. Overview: At loggerheads? Agricultural expansion, poverty reduction, and environment in the tropical forests. World Bank Policy Research Report, Washington, DC.

Coudel, E., Piketty, M., Gardner, T., Viana, C., Ferreira, J., Morello, T., Parry, L., Barlow, J., Antona, M., 2012. Environmental compliance in the Brazilian Amazon: exploring motivations and institutional conditions. ISEE 2012 Conference-Ecological Economics and Rio+ 20: Challenges and Contributions for a Green Economy, Available: http://www.isee2012.org (accessed 6.08.15.).

De Oliveira, J.A.P., 2008. Property rights: land conflicts and deforestation in the Eastern Amazon. Forest Policy Econ. 10,303-315.

Deininger, K., Feder, G., 2009. Land registration, governance, and development: evidence and implications for policy. World Bank Res. Obs. 24 (2), 233-266.

Duchelle, A.E., Cromberg, M., Gebara, M.F., Guerra, R., Melo, T., Larson, A., Cronkleton, P., Borner, J., Sills, E., Wunder, S., 2014. Linking forest tenure reform, environmental compliance, and incentives: lessons from REDD+ initiatives in the Brazilian Amazon. World Dev. 55,53-67.

Fearnside, P.M., 2001. Land-tenure issues as factors in environmental destruction in Brazilian Amazonia: the case of southern Pará. World Dev. 29,1361-1372.

Fearnside, P.M., 2003. Deforestation control in Mato Grosso: a new model for slowing the loss of Brazil's Amazon forest. AMBIO: J. Hum. Environ. 32(5), 343-345.

Geoghegan, J., Pritchard Jr., L., Ogneva-Himmelberger, Y., Chowdhury, R., Sanderson, S., Turner II, B., 1998. 'Socializing the pixel' and 'pixelizing the social'in land-use and land-cover change. In: In People and Pixels: Linking Remote Sensing and Social Science. National Academy Press, Washington, D.C, pp. 51-69.

Gibbs, H.K, Munger, J., L'Roe, J., Barreto, P., Pereira, R., Christie, M., Amaral, T., Walker, N.F., 2015. Did ranchers and slaughterhouses respond to zero-deforestation agreements in the Brazilian Amazon? Conserv. Lett. 9(1).

Gignoux, J., Macours, K, Wren-lewis, L., 2013. Evaluating the Impact of Land

Administration Programs on Agricultural Productivity and Rural Development. Inter-American Development Bank, Available: http://publications.iadb.org/ handle/11319/5810 (accessed 6.08.15.).

Godar, J., Tizado, E.J., Pokorny, B., 2012. Who is responsible for deforestation in the Amazon? A spatially explicit analysis along the Transamazon Highway in Brazil. For. Ecol. Manage. 267,58-73.

Godar, J., Gardner, T.A., Tizado, E.J., Pacheco, P., 2014. Actor-specific contributions to the deforestation slowdown in the Brazilian Amazon. Proc. Natl. Acad. Sci. 111,15591-15596.

INCRA, 2014. Instituto Nacional de Colonizacao e Reforma Agrária, Assentamentos Rurais, Available: http://acervofundiario.incra.gov.br (accessed 6.10.14.).

INPE, 2014. Instituto Nacional de Pesquisas Espaciais, Projeto de Monitoramento do Desmatamento na Amazonia Brasileira por Satélite (Available: http://www. obt.inpe.br/prodes (accessed 12.05.14.).

Jepson, W., 2006. Private agricultural colonization on a Brazilian frontier: 1970-1980. J. Hist. Geogr. 32, 839-863.

Larson, A.M., Brockhaus, M., Sunderlin, W.D., Duchelle, A., Babon, A., Dokken, T., Pham, T.T., Resosudarmo, I.A.P., Selaya, G., Awono, A., Huynh, T.-B., 2013. Land tenure and REDD+: the good, the bad and the ugly. Global Environ. Change 23, 678-689.

Liscow, Z.D., 2013. Do property rights promote investment but cause

deforestation? Quasi-experimental evidence from Nicaragua. J. Environ. Econ. Manage. 65, 241-261.

Liverman, D.M., Cuesta, R.M.R., 2008. Human interactions with the Eart system: people and pixels revisited. Earth Surf. Process. Landforms 33,1458-1471.

Michalski, F., Metzger, J.P., Peres, C.A., 2010. Rural property size drives patterns of upland and riparian forest retention in a tropical deforestation frontier. Global Environ. Change 20, 705-712.

Monteiro, T., 2015. CAR Fundiário vai agilizar processo de regularizacao fundiária no Estado. Agencia Pará [Online], Available: http://www.agenciapara.com.br/ noticia.asp?id_ver=109774 (accessed 6.08.15.).

Naughton-Treves, L., Wendland, K., 2014. Land tenure and tropical forest carbon management. World Development 55,1-6.

Oliveira, G., 2013. Land regularization in Brazil and the global land grab. Dev. Change 44, 261-283.

Portal Brasil, 2015. Cadastro Ambiental Rural é prorrogado até maio de 2016. Portal Brasil [Online], Available: http://www.brasil.gov.br/meio-ambiente/2015/05/ cadastro-ambiental-rural-e-prorrogado-ate-maio-de-2016 (accessed 6.08.15.).

Pacheco, P., Benatti, J.H., 2015. Tenure security and land appropriation under changing environmental governance in lowland Bolivia and Pará. Forests 6, 464-491.

Pacheco, P., 2009. Agrarian reform in the Brazilian Amazon: its implications for land distribution and deforestation. World Dev. 37,1337-1347.

Pacheco, P., 2012. Actor and frontier types in the Brazilian Amazon: assessing interactions and outcomes associated with frontier expansion. Geoforum 43, 864-874.

Paulino, E.T., 2014. The agricultural: environmental and socio-political

repercussions of Brazil's land governance system. Land Use Policy 36,134-144.

Republic of Brazil, 1980. Decree 84,685,6 May, Available: http://legis.senado.gov. br/legislacao/ListaNormas.action?numero=84685&tipo_ norma=DEC&data=19800506&link=s (accessed 6.11.15.).

Republic of Brazil, 2006. Law 11,326, 24 July, Available: http://www.planalto.gov. br/ccivil_03/_ato2004-2006/2006/lei/l11326.htm (accessed 6.11.15.).

Republic of Brazil, 2009a. Law 11,952, 25 June, Available: http://www.planalto.gov. br/ccivil_03/_ato2007-2010/2009/lei/l11952.htm (accessed 6.11.15.).

Republic of Brazil, 2009b. Decree 6,992, 28 October, Available: http://www. planalto.gov.br/ccivil_03/_ato2007-2010/2009/Decreto/D6992.htm (accessed 6.11.15.).

Republic of Brazil, 2012. Law 12,651, 25 May, Available: http://www.planalto.gov. br/ccivil_03/_ato2011-2014/2012/lei/l12651.htm (accessed 6.11.15).

Rajao, R., Vurdubakis, T., 2013. On the pragmatics of inscription: detecting deforestation in the Brazilian Amazon. Theory Cult. Soc. 30, 151-177.

Rajao, R., Azevedo, A., Stabile, M.C., 2012. Institutional subversion and deforestation: learning lessons from the system for the environmental licencing of rural properties in Mato Grosso. Public Admin. Dev. 32, 229-244.

Rajao, R., 2013. Representations and discourses: the role of local accounts and remote sensing in the formulation of Amazonia's environmental policy. Environ. Sci. Policy 30,60-71.

Reydon, B.P., Fernandes, V.B., Telles, T.S., 2015. Land tenure in Brazil: the question of regulation and governance. Land Use Policy 42, 509-516.

Richards, P.D., Vanwey, L., 2015. Farm-scale distribution of deforestation and remaining forest cover in Mato Grosso. Nat. Clim. Change.

Robinson, B.E., Holland, M.B., Naughton-Treves, L., 2014. Does secure land tenure save forests? A meta-analysis of the relationship between land tenure and tropical deforestation. Global Environ. Change 29, 281-293.

SIMLAM, 2008. Manual de Operacao de SIMLAM técnico. Secretaria de Estado de Meio Ambiente e Sustentabilidade (SEMA), Available: http://monitoramento. sema.pa.gov.br/simlam/index.htm (accessed 6.8.15., 6.11.15.).

SIMLAM-PA, 2013. Sistema Integrado de Monitoreamento e Licenciamento

Ambiental-Pará, Cadastro Ambiental Rural (CAR). http://monitoramento.sema. gov.br/simlam/.

State of Pará, Law 7,289, 24 July 2009; Available: http://www.pge.pa.gov.br/files/ u13/ARTlGO%2013%20-%20LEl%20n%C2%BA%207289%20OKK.pdf (accessed 6.1.15.).

State of Pará, Normative Instruction 37, 2 February 2010a. Available: http://www. semas.pa.gov.br/2010/02/02/10963/(accessed 6.11.15.).

State of Pará, Decree 2,135, 26 February 2010b. Available: http://www.pge.pa.gov. br/files/u13/ARTlGO%2011%20-%20DECRETO%20N.2135%20OKK.pdf (accessed 6.11.15.).

State ofPará, Decree 739, 28 June 2013. Available: http://www.semas.pa.gov.br/ 2013/06/28/d-e-c-r-e-t-o-no-739-de-29-de-maio-de-2013/(accessed 6.11.15.).

State ofPará, Decree 1,148,17July 2008. Available: http://www.sema.pa.gov.br/ 2008/07/17/9687/(accessed 6.11.15.).

Schmink, M., 1982. Land conflicts in Amazonia. Am. Ethnol. 9,341-357.

Simmons, C.S., 2005. Territorializing land conflict: space, place, and contentious politics in the Brazilian Amazon. GeoJournal 64,307-317.

Soares-Filho, B., Rajao, R., Macedo, M., Carneiro, A., Costa, W., Coe, M., Rodrigues, H., Alencar, A., 2014. Cracking Brazil's forest code. Science 344,363-364.

Stickler, C.M., Nepstad, D.C., Azevedo, A.A., Mcgrath, D.G., 2013. Defending public interests in private lands: compliance, costs and potential environmental consequences of the Brazilian Forest Code in Mato Grosso. Philos. Trans. R. Soc. B: Biol. Sci. 368, 20120160.

Terra Legal, 2015. Available: http://i3geo.mda.gov.br/i3geo/datadownload.htm (accessed 11.2.15.).

Wright, A.L., Wolford, W., 2003. To Inherit the Earth: the Landless Movement and the Struggle for a New Brazil. Food First Books.